|
PureBytes Links
Trading Reference Links
|
> Baloney....T3 is very slow reacting.......HIGH LAG !
Not true.
> The lookback length must be approximately HALF of that of the one
> used in a JMA plot to come even close.
That's because the "length" inputs do not mean the same thing on the
two functions. With the same inputs T3 has more lag than JMA, but
that's because it's smoothing a lot more. Set them to whatever
lengths it takes to get about the same lag, and you'll see they are
very comparable. Filters like this are a lag/noise tradeoff; with
less lag you get more noise, with less noise you get more lag. With
the same lag the two averages have about the same amount of noise.
Tim Tillson (not Tillman) is a friend of mine -- I've known him for
over 20 years. I've also spoken to Mark Jurik about T3, and listened
in when the two of them had some pretty interesting (& deep!)
discussions at a recent MTA meeting. (It was the first time they'd
met face-to-face, in spite of years of friendly rivalry.) Mark
freely admits that T3 blew the doors off his previous version of his
AMA, and that spurred him to do the long & intense research it took
to create JMA. Mark also says that T3 is very comparable to JMA, but
JMA follows gaps faster and has less overshoot. Look at the charts
on http://www.jurikres.com/catalog/ms_ama.htm and you can see this
for yourself.
Tim says he has an average that's about as good as JMA, but he
prefers to keep that one proprietary. Mark is coming out with JMA2
next month, which he says will cut the lag even more. Then he says
he's quitting the "moving average escalation wars." These things are
so close to the best that can theoretically be done, it takes
thousands of hours of work to come up with a tiny improvement in
performance. Mark is going to focus his efforts elsewhere.
So, **according to Mark Jurik**, JMA is very slightly better than T3.
T3 is free. Take your pick.
Gary
|